spalart shur correction implementation

[romoli]

Hello,

I am using turbomachinery with CS_TURBOMACHINERY_FROZEN model and with the spalart shur correction activated.

I looked at the formulation of this correction from the thesis of M. De Laage De Meux and also by looking at the source code.

And it bring me out some questions.

If I understand ( which is certainly a first very opimistic hypothesis ) the correction factor include a term that depends on the rotation matrix :


The code says that when coriolis forces are activated, the rotation vector is used to compute the rotation matrix for the above mentioned term.

My interrogation is then when not using the coriolis forces but when using turbomachinery frozen model, does the rotation matrix term appears somewhere or not ? Why there is no part of the code testing if we are using turbomachinery and getting information of rotation vector to compute the term linked to the rotation matrix ?

Is this a relative / absolute reference frame issue that made the missing term not necessary ?

Thank you for your light ...

edit -> I am using code_saturne v3.2.1

[Yvan Fournier]

Hello,

I'll suggest to Benoît to look at your post and give a more detailed answer, but one aspect which may help you navigate the code is that with the turbomachinery option, only part of the domain is rotating, while with the old method, coupling 2 domains, one of the domains is rotating.

Still, I will check with Benoît next week (I am out of office this week) to make sure no term is missing.

Regards,

Yvan

[romoli]

Thank you for considering my interrogations.

I just can make the precision that I am using the turbomachinery im my whole domain such that all the fluid is in rotation.

Regards.

[BLM]

Hi,

In the Spalart-Shur correction, the "Coriolis-type" term you mention comes from the material derivative of the strain rate tensor (this term is used to "measure" the curvature of the streamlines and, since it is not an objective quantity, it generates Coriolis-type term in a rotating reference frame).

However, with the CS_TURBOMACHINERY_FROZEN option, all quantities are related to the absolute reference frame. The unique exception is the mass flux (rho.u)S which is related to the relative velocity. It is a choice of implementation, for stability reasons, from one part, and by analogy with the ALE formulation, from the other part.

In the implementation of the Spalart-Shur correction, the mass flux is not used to compute the material derivative of the strain rate tensor (as it should be, in a finite volume spirit). We directly contract the gradient with the absolute velocity (in a source term fashion...). As a result, there is no need to the Coriolis counter-part of the material derivative, using the Spalart-Shur model with the CS_TURBOMACHINERY_FROZEN option.

Note that in RSM models equation, ther is 2 terms that generate the same Coriolis-type term : the turbulent production and the material derivative of the Reynolds stresses. As a result, the Coriolis-type term appears with a "2" factor in this equation in a rotating frame of reference (a "2" factor appears also in the basic Coriolis term, in the momentum equation). With the CS_TURBOMACHINERY_FROZEN model, there is only a "1" factor for this term because the material derivative is computed using the mass flux (related to the relative frame of reference, as mentioned above) while the production is computed using the absolute velocity, as others source terms.

I hope all my explanations are not too confusing and I apology for the delay in my response.

See you

[romoli]

Hello,

Thank you very much for your detailed explanations.

I have clearer ideas now on the use of correction with turbomachinery.
For the RSM models remark is still quite confusing for me so I have to dig into it to get clear.

My RSM experience (with SSG model) is until now quite disappointing since i expected that it would give better results than kw with correction.
In fact it did get better in a first period during the simulation but then the solution evolves and finally get worst than the solution obtained by kw with correction.

I guess it is a mesh / scheme / convergence issue.

Thank you again for spending time giving me an answer.